Think or Remember: The Exchange Rate of Agentic Inference
cs.CL, cs.AI
Submitted: 2026-11-07
Updated: 2026-11-07
Terminology
Sources
- State commitment learning: training language models to distinguish computation from memory
- Continuous Context Management
- Scaling Laws for Neural Language Models
- Self-Compacting Language Model Agents
- Plans Fade Fast: Why Context Management Is Load-Bearing for LLM Agents
- CliffCompaction: Cost-Efficient Compaction for Long-Horizon Coding Agents
- KV-streams for Efficient Compaction in Agentic Reinforcement Learning
- Beyond Token Savings: A Systematic Study of Context Compression in LLM Agents
- Beyond Compaction: Structured Context Eviction for Long-Horizon Agents
- Context Language Models
- StateComp: Learning When to Compress History in Long Horizon Agents
- When Can Agents Forget Their Reasoning? ICLR for Long-Horizon Agent Context Compression
- AutoCompact: Learning When to Compact Context in Long-Horizon Coding Agents
Related papers
- ALARM: Audio-Language Alignment for Reasoning Models
- ThinkSum: Probabilistic reasoning over sets using large language models
- Coherence boosting: When your pretrained language model is not paying enough attention
- From Facts to Insights: A Persona-Driven Dual Memory Framework and Dataset for Role-Playing Agents
- Improving General Role-Playing Agents via Psychology-Grounded Reasoning and Role-Aware Policy Optimization
- The Art of Asking: Multilingual Prompt Optimization for Synthetic Data